Research on Risk Identification and Quantification for Heavy-Duty Trucks Based on Multi-Feature Analysis

To correct the habitual dangerous driving characteristics of drivers and address the existing shortcomings in accurately and comprehensively analyzing dangerous driving behaviors, this paper proposes a method for identifying and quantifying drivers’ on-road driving risks in heavy-duty trucks. Firstly, natural driving experiments were designed, during which multi-modal driving data from heavy-duty trucks were collected using an inertial navigation recorder, eye-tracking recorder, and electrocardiograph recorder. Secondly, based on feature threshold calculation, dangerous driving behaviors were extracted. Then, combining driver feature extraction and the frequency of dangerous driving behaviors, a classification of driving styles for drivers is achieved through the combined approach of SOM and K-means clustering. Finally, a quantitative assessment model for driving styles was developed using the CRITIC weighting method. The research findings hold promise for providing guidance and theoretical support for subsequent driver risk assessments and targeted traffic management.